Table 7 Suggestions for further methodological research and justification

Challenge 1—Capturing non-health outcomes and societal intervention costs

 1. Explore methods for consulting stakeholders on the appropriate perspective to take (e.g., public sector, societal) and the range of appropriate outcomes and costs, particularly under the societal perspective

Models operationalising the societal perspective were generally limited in terms of the range of societal outcomes and costs incorporated; see Table 2 and “Capturing non-health outcomes and societal intervention costs” section. The appropriate perspective for the evaluation would depend on the range of outcomes prioritised by decision stakeholders [39]

 2. Explore methods and data sources for incorporating balanced sets of outcomes and intervention costs under the societal perspective

Balanced incorporation of non-health outcomes and societal intervention costs was achieved by only two of 15 models shown in Table 2; see “Capturing non-health outcomes and societal intervention costs” section. Imbalanced incorporation would risk over- or under-estimating the cost-effectiveness of the intervention

 3. Explore methods to account for sector-specific productive efficiencies under the societal perspective and to assess the relevance of established/possible cost-effectiveness thresholds [108]

Accounting for the intersectoral differences in cost-effectiveness thresholds (i.e., productive efficiencies) would have changed the final decision in several models [94, 99, 100]

Challenge 2—Considering heterogeneity and dynamic complexity

 1. Explore methods and data sources for incorporating variables that depict geriatric health and falls risk variations within the same age and sex groups and over time, such as the continuous, multivariate frailty index [152]

Models were limited in terms of incorporating subgroup delineators beyond age and sex (Table 3) and time-variant falls risk factors beyond age and falls history (Table 4). The multivariate frailty index is suggested as a variable that can capture the multidimensional nature of changes to geriatric health and falls risk

 2. Explore the impact on intervention rankings of the choice in the main decision metric between cost-per-unit ratio and aggregate outcome [104]

Several models evaluated interventions targeting heterogeneously sized subgroups then compared the resulting ICERs only. This may have introduced misleading interpretations of economic outcomes: see the last paragraph of “Heterogeneity” section

 3. Explore methods and data sources for characterising the heterogeneity in intervention efficacy, cost and implementation level

Heterogeneities in intervention efficacy and cost were characterised by only one model each (“Heterogeneity” section, 3rd paragraph). Heterogeneities in intervention access, compliance, and sustainability were likewise highly limited (Table 5)

 4. Explore the feasibility of developing individual-level simulation to capture the age-related progression in falls risk and other dynamic patterns in geriatric health aspects (e.g., progressions in comorbidity care costs)

Tunnel states were not described for the 13 cohort-level Markov models in Table 4. Individual-level models are likely better suited to characterise the age-related falls risk progression. Health utilities and comorbidity costs progressed by age groups only. Here again, individual-level simulation can capture the annual progressions and variations by geriatric health variables (e.g., frailty, functional status)

 5. Explore methods for modelling: (i) periodic falls risk screening to allow dynamic variation in the proactive intervention pathway (5); and (ii) access to reactive pathway after a serious falls incidence

No model incorporated repeated falls/fracture risk screening to reassess the need for proactive intervention access. Only one model shifted individuals to the reactive pathway once a fracture occurred (“Dynamic complexity” section, 3rd paragraph)

 6. Explore methods for modelling incoming cohorts of newly eligible persons to characterise the dynamic target population size and capacity implications

Models mentioned the non-incorporation of incoming cohorts as a limitation that underestimated the total intervention costs and benefits (“Dynamic complexity” section, 5th paragraph). No model considered capacity implication, most affected by incoming cohorts who generate sustained intervention need (“Considering theories of human behaviour and implementation” section, 5th paragraph)

Challenge 3—Considering theories of human behaviour and implementation

 1. Explore methods and data sources for incorporating individual- and social-level variables that influence health behaviour and intervention supply/demand

No model directly parameterised psychological and social causal mechanisms based on individual and social behavioural theories (“Considering theories of human behaviour and implementation” section, 1st paragraph). In Table 5, only one model characterised the long-term variation in demand persistence

 2. Explore methods for distinguishing between supply- and demand-side implementation factors and evidence sources for long-term sustainability of interventions

Only Turner et al. [124] distinguished between demand-side uptake and supply-side adoption as determinants of initial access (“Considering theories of human behaviour and implementation” section, 2nd paragraph). Sustainability parameters relied extensively on assumptions (Table 5; “Considering theories of human behaviour and implementation” section, 2nd paragraph)

 3. Explore the feasibility of conducting value of implementation analyses as alternative scenarios of implementation strategies with aggregate monetary outcomes to estimate the willingness to pay

Models often assessed the variations in implementation levels under DSA (i.e., to assess parameter uncertainty) rather than scenario analysis (“Considering theories of human behaviour and implementation” section, 3rd paragraph). Cost-per-unit ratios may poorly indicate the impact of implementation change (“Considering theories of human behaviour and implementation” section, 3rd paragraph)

 4. Explore the feasibility of developing models that explicitly incorporate capacity constraints, such as discrete events simulation [153]

No model considered the capacity or budget implications of their interventions. This resulted in misleading outcomes; an example is given in “Considering theories of human behaviour and implementation” section, 5th paragraph

Challenge 4—Considering issues of equity

 1. Explore methods for consulting stakeholders to identify relevant social and health severity delineators

Table 6 shows that few models incorporated vulnerable subgroups; ethnicity was the only social delineator of equity relevance. Consulting the stakeholders is facilitates equity analyses relevant to the specific settings (“Considering issues of equity” section) [39]

 2. Explore methods and data sources for modelling causal mechanisms behind vulnerable subgroups’ reduced capacity to benefit

Models in Table 6 did not fully account for causal mechanisms of reduced capacity to benefit: e.g., BODE3 models did not parameterise heterogeneous intervention efficacy and access by ethnicity, precluding analyses of the double jeopardy problem (“Considering issues of equity” section, 2nd paragraph)

 3. Explore methods for equity analysis that assesses the equity-efficiency trade-off under alternative intervention strategies, such as DCEA [82]

No model evaluated alternative strategies that prioritised the vulnerable subgroups and then estimated the efficiency-equity trade-off (“Considering issues of equity” section, 6th paragraph)